Modified inter prediction H.264 video encoding for maritime surveillance

• Main Author: Mohd. Razif, Mohd. Azraf
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://eprints.utm.my/id/eprint/78830/1/MohdAzrafMohdMFKE2016.pdf; http://eprints.utm.my/id/eprint/78830/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:109570

Video compression has evolved since it is first being standardized. The most popular CODEC, H.264 can compress video effectively according to the quality that is required. This is due to the motion estimation (ME) process that has impressive features like variable block sizes varying from 4×4 to 16×16 and quarter pixel motion compensation. However, the disadvantage of H.264 is that, it is very complex and impractical for hardware implementation. Many efforts have been made to produce low complexity encoding by compromising on the bitrate and decoded quality. Two notable methods are Fast Search Mode and Early Termination. In Early Termination concept, the encoder does not have to perform ME on every macroblock for every block size. If certain criteria are reached, the process could be terminated and the Mode Decision could select the best block size much faster. This project proposes on using background subtraction to maximize the Early Termination process. When recording using static camera, the background remains the same for a long period of time where most macroblocks will produce minimum residual. Thus in this thesis, the ME process for the background macroblock is terminated much earlier using the maximum 16×16 macroblock size. The accuracy of the background segmentation for maritime surveillance video case study is 88.43% and the true foreground rate is at 41.74%. The proposed encoder manages to reduce 73.5% of the encoding time and 80.5% of the encoder complexity. The bitrate of the output is also reduced, in the range of 20%, compared to the H.264 baseline encoder. The results show that the proposed method achieves the objectives of improving the compression rate and the encoding time.